IDL: Don't lose nullable flag when "preprocessing" type

IDL: Don't lose nullable flag when "preprocessing" type

The preprocessing converts any enumeration type into IdlType('DOMString'),
including nullable enumeration types, which is wrong. It should convert an
nullable enumeration type into IdlNullableType(IdlType('DOMString')).

This change affects the code generation for the following methods and
attributes:

  object? WebGLRenderingContextBase.getExtension()
  VendorIDSource? BluetoothDevice.vendorIDSource [can return null]
  RTCSdpType? RTCSessionDescription.type [can be set to and return null]

BUG=451602
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=190259

[Jens Widell, 2015-02-16 14:33:53]

jl@opera.com changed reviewers:
+ haraken@chromium.org, scheib@chromium.org

[Jens Widell, 2015-02-16 14:33:54]

PTAL

https://codereview.chromium.org/928233002/diff/40001/Source/bindings/scripts/...
File Source/bindings/scripts/v8_types.py (left):

https://codereview.chromium.org/928233002/diff/40001/Source/bindings/scripts/...
Source/bindings/scripts/v8_types.py:677: if (idl_type.name in ['Any', 'Object']
or idl_type.is_callback_function):
Looking at idl_type.name here was wrong; it would be AnyOrNull/ObjectOrNull for
a nullable type (and was in the case of
WebGLRenderingContextBase.getExtension()'s "object?" return value.)

With the other change in this function this code would now have worked, but I
changed it anyway, since I think it's better to look at base_type in most cases;
using name is equivalent except it invites this kind of error.

https://codereview.chromium.org/928233002/diff/40001/Source/core/html/canvas/...
File Source/core/html/canvas/WebGLRenderingContextBase.h (left):

https://codereview.chromium.org/928233002/diff/40001/Source/core/html/canvas/...
Source/core/html/canvas/WebGLRenderingContextBase.h:202:
PassRefPtrWillBeRawPtr<WebGLExtension> getExtension(const String& name);
This signature worked only by mistake. A method returning "object" is expected
to return ScriptValue, and the binding code will be

  v8SetReturnValue(info, <expression>.v8Value());

which compiles when <expression> is a ScriptValue, but not when it is what this
function returns.

However, this method was declared to return "object?", and there was a bug that
then disabled the conversion from "object" to "ScriptValue" (because the type
was nullable), the binding code ended up being

  v8SetReturnValue(info, <expression>);

which compiles fine when <expression> is what this function returns.

[haraken, 2015-02-16 14:42:00]

LGTM

[Jens Widell, 2015-02-16 15:54:48]

The CQ bit was checked by jl@opera.com

[commit-bot: I haz the power, 2015-02-16 15:55:40]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/928233002/40001

[commit-bot: I haz the power, 2015-02-16 16:56:40]

Committed patchset #3 (id:40001) as
https://src.chromium.org/viewvc/blink?view=rev&revision=190259